Baseball Prospectus - Chat with Harry Pavlidis

Chat: Harry Pavlidis

Welcome to Baseball Prospectus' Thursday April 30, 2015 1:00 PM ET chat session with Harry Pavlidis.

Harry is Baseball Prospectus' director of technology. Ask him about BP's brand new Deserved Run Average, the most perfect pitching metric ever devised. Or about the latest in baseball analytics, or about the Cubs, or about anything else in the whole world.

Harry Pavlidis: Happy May Day Eve! Hopefully Sam Malone will bring you all the gifts you wished for.

gilpdawg (Ohio): Nice job on DRA....it looks promising. Now that the buttkissing is out of the way, is the injury database done for good as far as new updates, or is there a plan to bring it back at some point?

Harry Pavlidis: Thank you for the totally sincere feedback :)
But, sorry, no current plans to make it come back, but I imagine if enough people ask ...

weskelton (New Jersey): Hey Harry, Over the last week or so, we've started to see StatCast and the potential of the data being gathered by SportsVision. What do you hear about the availability of the underlying data to you (BP) and to us (the general public)? Thanks, -Bill S.

Harry Pavlidis: MLBAM is planning to keep it/make it available. We look forward to getting into it. A few of us here have experience with the Trackman data (batted ball, pitcher stuff) already so we're excited to have that stuff available in the public domain.

Doug (Rochester, NY): I expected the next big pitching metric breakthrough to come via Statcast, if only because I thought any improvements that could be made to pitching metrics without new data would already have been made. DRA doesn't seem like it needed to have been made in 2015; most of the data looks like it has been available for several years. Why now? What has changed that made DRA possible?

Harry Pavlidis: DRA is more about the modeling/process than the data itself. Also, some of the metrics (framing etc) are relatively new, but, yes, for the most part, the ingredients existed for some time. What made DRA possible was our comfort level with complex modeling and its application to baseball.

weskelton (New Jersey): any idea on the timing of when it would be available?

Harry Pavlidis: Soon, I hope, but no specific timelines.

Daniel Schoenfeld (Evanston, IL): How strong of a correlation would you guess exists between TRAA and a pitcher's time to home plate?

Harry Pavlidis: quite a bit. It's a combination of time to home, varying time from set to go, pick off moves, looks over. But, if you're slow (or fast) to home that's probably the trump card.

doug (the moooooon): This isn't a defense-independent metric, right? If correlating well to RA9 is the goal, why not use RA9? It feels like this is basically just a component-adjusted RA9

Harry Pavlidis: the idea here is to tell us what the pitchers' RA9 _should have_ been, not what it was. Correlating at the population level is expected/desired, but the individual differences are what DRA is trying to capture.

Dan (OR): Because it adjusts for context so heavily (including base-out state), should we view this measure as less predictive than some of the current existing measures?

Harry Pavlidis: Yes, while it does have predictive power it falls short of cFIP. We still suggest using PECOTA for forecasting purposes.

ADprospectus (Atlanta): Novice question about DRA: I understand that the purpose of DRA is to more precisely account for the pitcher's responsibility for runs the opposing team scores, but I don't understand why the quality of DRA is graded by its ability to track RA/9. If RA/9 is the benchmark or standard against which things like DRA, FIP, and SIERA are measured, why shouldn't we just look to RA/9? Conversely, if we shouldn't just look to RA/9 to understand the quality of past pitcher performance, why grade DRA against RA/9?

Harry Pavlidis: hopefully my answer to an earlier question covers this--if not please follow up for additional clarification!

Typical guy (Internet): TL:DR Thanks ill hang up and listen.

Harry Pavlidis: you forgot "long time caller first time listener" which is how the internet really works btw

steelydanu (Los Angeles): Hello, I have Lucas Giolito on my Minor roster. I have not heard anything on him in terms of making starts this year. Is WAS really being that conservative with him?

Harry Pavlidis: he's in extended ST, he'll be up in full season ball this upcoming week, per reports from MASN.

Pelecos (Granville): Hey Harry, is there any new data that MLB stat cast will give the public that could help enhance DRA in the future? What are some of the shortcomings of the statistic? Thanks

Harry Pavlidis: Batted ball velocities and angles will be very helpful.
I'd say the 'shortcoming' is complexity. Right now we are still working on getting all the bugs smoothed out, and investigating some potential sources of error. So, on one hand complexity allows us to wring out more information from existing data, but on the other hand it creates more work. Hence the team approach, which is working out great. Right now the gang is looking into some data bugs and researching a possible bias (over-crediting the best pitchers) pointed out by an astute reader.

steelydanu (Los Angeles): hello, Who would you choose as #1's: R. Lopez WAS or A. Reyes STL?

Harry Pavlidis: I love both. I think Reyes, because he's further along. But right now I'd be cautious tagging either with the #1 label.

doug (texas): If the goal of DRA is to strip out sequencing, the contributions of other players, the atmospheric conditions, etc., and get to what should have happened, how is that different from a true-talent measurement?

Harry Pavlidis: there's some luck left in there. This metric is designed first to describe what happened, with the fairest possible accounting of responsibility we can muster.

bargain (Denver): What was the biggest headache from the months of daily stat creation? Was there one thing that took excruciatingly longer than anticipated?

Harry Pavlidis: nothing was painful. It's actually fun doing this stuff. We're still trying to get all the historical data perfect, so that's a bit of a frustration. Our next trick will be trying to slice it up into time segments within seasons, so I guess you can say that took longer than planned since we haven't done it yet.

Daniel Schoenfeld (Evanston, IL): How would you recommend learning about mixed models and their utilization?

Harry Pavlidis: besides hiring Jonathan Judge? Great question, for starters I'd check out some of the resources in the CSAA article (our first foray into MM), and bug Judge on twitter @bachlaw :)

Alex (Anaheim): Who has the most impressive bullpen so far?

Harry Pavlidis: 1990 Reds

AA (FL): What's your opinion of Jeff Hoffman? Potential top 25 prospect heading into next year?

Harry Pavlidis: if he pitches, sure. He's highly regarded but the question marks are obviously waiting for the eraser known as "actually playing baseball" in the very near future.

3FingerTimm (California): Jeremy Null? Someone to pay attention to or just on a ridiculous hot streak to start the season?

Harry Pavlidis: maybe both. Tall dude, so the "click" can happen. Give it more time, but pay attention.

Cal Guy (Cal): Harry, Do you think something has clicked for Tim Beckham, and do you think he becomes the starting 2B at some point?

Harry Pavlidis: talking about clicking ... maybe. He may also be a little lucky/hot. So, see answer above ;)

Harry Pavlidis: Cheers all, enjoy some baseball and let's meet back here in a few weeks to talk ... maybe about game calling.